Value-added trade and international shock transmission in ...
Transcript of Value-added trade and international shock transmission in ...
1
Value-added trade and international shock transmission in a New
Keynesian model of production fragmentation
Chin-Yoong Wonga,*
Yoke-Kee Enga
aDepartment of Economics, Faculty of Business and Finance, Universiti Tunku Abdul Rahman, Jalan
Universiti, Bandar Barat 31900 Kampar, Perak, Malaysia.
* Corresponding author. Tel: 6054688888, ext. 1035. Fax: 6054667407. Email address:
Abstract
This paper proposes a New Keynesian model of production fragmentation that allows the
decomposition of trade into value added components compatible with the most
comprehensive definition of value-added trade in the empirical trade literature. In so doing, a
country’s position in the value chains of production can be identified transparently. The
model is then estimated on ten East Asian economies by using Bayesian method. We find that
countries vertically bound but specialized in different value chains tend to respond uniformly
towards supply and demand shocks. The magnitude of the response is also shaped by the
country upstreamness in production sharing.
Keywords: Value-added trade; Production fragmentation; Vertical specialization; New
Keynesian; Bayesian estimation
JEL classification: F14, F41, F44,
2
1. Introduction
One of the defining characteristics of today’s international trade is the heavy flow of
parts and components between countries. Trade barrier reduction in the 1980s accelerated the
growth of world trade, effected the integration of most of the developing countries into the
world economy, and caused a reorganization of the production structure. Productions are
vertically sliced and fragmented across countries with different specializations, resulting in
multiple back-and-forth trades in intermediate goods until a final product is assembled
(Feenstra 1998).
In the presence of foreign value added, gross values of exports and imports are no
longer an appropriate measure of a country’s integration into the world economy. Who
produces for whom in the world economy, to quote Daudin et al. (2011), becomes a puzzle
waiting to be solved. In response to the measurement challenge, literature has been
blossoming ever since the pioneering Hummels et al. (2001) that first defined the meaning of
vertical specialization, which many subsequent empirical works have been adopted and
improved1.
However, the importance of tracing the sources and uses of value added that are
embodied in trade goes beyond the measurement challenge. When countries are tied together
through global production sharing, how does shock propagate across chains of production
and thus countries? In terms of magnitude of the response, does it matter whether a country
lies in the upstream or the downstream of production in the face of supply and demand
shocks? Understanding the macroeconomic interdependence between countries participating
in production sharing is fundamentally critical for other more important macroeconomic
questions such as the design of optimal monetary coordination.
1 According to Hummels et al. (2001), vertical specialization occurs when goods are produced in multiple
sequential production stages involving two or more countries. At least one country must use imported goods in
its chain of processing, and some of the resulting output must be exported either as intermediate or final goods.
3
To take up this challenge, this paper blends the recent and actively growing literature
on the measurement of the value added in trade by using the global input-output framework
with the workhorse two-country dynamic New Keynesian model expanded to feature three
sequential processing stages. Questions such as whether a country participates in production
fragmentation or whether a participating country specializes at the upstream, midstream or
downstream stage of the production process are answered endogenously by the data using
Bayesian inference.
Building on the work by Johnson and Noguera (2012a, 2012b), Daudin et al. (2011),
Hummels et al. (2001), and, particularly, Koopman et al. (2010), we decompose the gross
exports, generated by using the Bayesian estimated model, into domestic and foreign value
added that are compatible with the most comprehensive definition of value added in gross
exports available in the empirical literature. In so doing, trade structure of the model becomes
transparent as we know where the value was added and how the value added is used. In
addition, in the spirit of Johnson and Noguera (2012a, 2012b), we identify the extent of
bilateral production fragmentation, and, similar to Koopman et al. (2010), we have
constructed an index that gauges the country upstreamness in production sharing.
Yi (2003) preceded our effort in modeling three production sequences through which
the final output is produced. However, what distinguishes our model from his dynamic
Ricardian trade model with nontradable final goods is that the outputs of all stages in our
model are tradable. This simple innovation is fundamental in quantifying the domestic value
added embodied in the export of intermediates shipped back to the source for downstream
production and in accounting for the foreign value added in the final domestic product
exports. We show that these are two components in value added that are important in
identifying the country’s position in the value chains.
4
Next, we use the framework to shed light on international shock transmission. Our
results indicate that whether favorable total factor productivity shock originates from an
upstream or downstream country, it tends to lift up the aggregate value added of countries
tied together by production sharing. In other words, vertical linkage in production has
synchronized the responses of participating countries toward shocks (di Giovanni and
Levchenko, 2010). The responses of an upstream country are magnified when its shock
correlates with the shock originated in a downstream country. This finding corroborates the
recent empirical finding regarding the role of vertical specialization in the great trade collapse
(Bems et al., 2010; Levchenko et al., 2010).
2. A New Keynesian model of production fragmentation: Key ingredients
In this section, we outline a two-country New Keynesian model of production
fragmentation. We added two innovations to the otherwise standard New Keynesian model.
First, final output is processed through three sequential stages. The upstream firms combine
labor services and capital in Cobb-Douglas production technology to produce materials that
are partially exported abroad for subsequent processing, with the remaining materials used
for local midstream remanufacturing in conjunction with the imported intermediate inputs. A
fraction of the remanufactured intermediate goods would then be re-exported for final
assembly in downstream production. The assembled final goods are consumed and invested
locally or exported. Invested final goods constitute the capital inputs for upstream production,
which later become the intermediates for subsequent processing. As such, we have outlined a
model of production fragmentation with “intermediate loops”. Figure 1 provides illustrations
on this production and trade structure.
[INSERT FIGURE 1 HERE]
5
Whereas a typical New Keynesian model with two stages of production can only
account for import of intermediates for final assembly and export in sequence, our model
simultaneously embraces the equally important sequential import and export of intermediate
inputs. With this parsimonious model, we can comprehensively trace the cross-border flow of
value added in the spirit of Koopman et al. (2010).
The second innovation is the assumption of trade invoicing in U.S. dollars as most
international trade is conducted in U.S. currency (Goldberg and Tille, 2008). This innovation
contradicts the typical practice in New Open-Economy Macroeconomics that assumes either
producer or local currency pricing.
Given the dollar price of export, depreciation against the U.S. dollar increases unit
export price in local currency. Although nominal depreciation translates into a higher local
price of imported intermediate inputs, the expanding export revenue helps firms absorb the
negative impact of depreciation on markup without the need to proportionally raise the output
price for the home and export market. As a result, the dollar pricing mechanism lies between
a zero exchange rate pass-through in local currency pricing mechanism and a complete
exchange rate pass-through into output price in producer currency pricing mechanism, while
retaining close movement between the nominal exchange rate and the terms of trade.
Below, we elaborate on our discussion of the model.
2.1. Chains of production from upstream to midstream and downstream
A unit mass of competitive firms at upstream production has access to Cobb-Douglas
production technology of (1) that uses plant-specific capital �����, � ∈ �, and a continuum of
differentiated labors � = � �(�)��,�����,� ���∈� ���,���,���� to produce plant-specific materials ����
at date t.
6
���� = ��������� !(�)��! (1)
where "� is a first-order autoregressive Hicks-neutral total factor productivity (TFP) shock.
The upstream firms can only alter its capital over time by varying the rate of investment #��
that comes with a cost S�#�� #����$ . Capital stock accumulation evolves according to the form
in Mandelman et al. (2011)
��� = (1 − ')����� + u�� #�(�) )1 − Ψ* +u���, ����(�)u�,��(�) - + .�,��(�).���, ����(�) − Λ-*/ (2)
where u�� is investment-specific technology (IST) shock, and follows first-order
autoregressive process. The parameter Ψ denotes investment adjustment cost, and Λ
determines how forward-looking the investment decision is.
The upstream firm thus optimally chooses the path of ���, �, and #�� to minimize the
cost of production (01,� + ')��� + 2�� subject to output net of investment adjustment cost,
Φ��� )��������� !(�)��! − Ψ* u���� #���� 3 .�,��4.���, ����4 − Λ5* = 0/ , where Φ��� is Lagrangian
multiplier which we define as the real marginal cost for upstream firm.
2� = � 2�(�)��,�����,� ���∈� ���,���,���� refers to the real wage and 01,� is the real return on capital.
Both correspond to respective marginal productivity. 78,� denotes the wage elasticity of the
demand for labor �. For the sake of simplicity, we assume that the market for upstream goods
is tightly competitive. The elasticity of substitution between varieties is thus close to infinity,
and as a consequence, price approximates real marginal cost. The output are used as
intermediates for either domestic or foreign midstream production, ���� ≡ ��:��: + ��:��;. The
marginal product of capital stock and labor, and the intertemporal optimal investment
decision are derived as
7
����� = 3 �<=,�5 >����� (3)
� = + �?�- (1 − >)Φ����� (4)
Φ�� 3 .�,��4.���, ����4 − Λ5 = Φ��@� )3.�A�, ��A�4.�,��4 − Λ5 3.�A�, ��A�4
.�,��4 5 − �* 3.�A�, ��A�4.�,��4 − Λ5*/ (5)
Equations (3) and (4) combined with production function (1) give us the real marginal cost of
upstream firm.
��� = +<=,�4 @B-C(?�)��CDE�!C(��!)��C (6)
A mass continuum of midstream monopolistically competitive firm �, � ∈ �, imports
c.i.f upstream goods F�;,��: of plant � for remanufacture in combination with local inputs ��:,��:
using CES production technology as in equation (7)
�*�� = �(1 − G*)�H���:,��: H��H + (G*)�H�F�;,��: H��H � HH�� (7)
whereF�;,��: = 3 F�;,��: (�)�I����I� ���∈J 5 �I��I��� and ��:,��: = 3 ��:,��: (�)�I����I� ���∈J 5 �I��I���
. The optimal
demand function for the � varieties of F�;,��: (�) is �L�;,�� (�) L�;,��$ �MI�F�;,��:, and of ��:,��: (�) is
�L�:,�� (�) L�:,��$ �MI���:,��:. L�:,��
and L�;,�� are the domestic price of local and imported
materials, respectively. 7*� > 1 is the time-varying demand elasticity. The parameter G*
indicates the share of imported intermediates in midstream production, and the parameter
O > 0 denotes the elasticity of substitution between home and imported intermediate inputs.
Midstream firm � minimizes the cost function L�:,�� ��:,��: + L�;,�� F�;,��: subject to
Φ*�� P�*�� − �(1 − G*)�H���;,��: H��H + G* �H�F�;,��: H��H � HH��Q, where Φ*�� is real marginal cost of
8
midstream firm to obtain the following optimal demand schedules for domestic and foreign
inputs
��:,�� = (1 − G*) RS�T,�4S�,�4 U�V �*�� (8)
F�;,��: = G* RS�W,�4S�,�4 U�V �*�� (9)
By substituting the first-order condition for ��:,��: and F�;,��:
, we can obtain the real marginal
cost for midstream firm.
Φ*�� = L��� = X(1 − G*)�L�:,�� ��V + G*�L�;,�� ��VY ���H (10)
Note thatL��� is the flexible production-based producer price for midstream production. The
market is cleared by the demand from home and foreign downstream firm, �*�� ≡ �*:,��: + �*:,��;.
Lastly at downstream production, a continuum of monopolistically competitive final-
good producers � of measure � combines a variety of home �*:,��: and imported intermediate
goods F*;,��:using the following CES technology to produce consumer goods.
�Z�� = �(1 − GZ)�H��*:,��: H��H + GZ �H�F*;,��: H��H � HH�� (11)
where �*:,��: = + �*:,��: (�)(M[���) M[�⁄ ���∈J - �[��[��� and F*;,��: = + F*;,��: (�)(M[���) M[�⁄ ���∈J - �[��[���
.
The demand schedules for � varieties of domestic and imported intermediates take the form
�*:,��: (�) = �L*:,�� (�) L*:,��$ �M[��*:,��: and F*;,��: (�) = �L*;,�� (�) L*;,��$ �M[�F*;,��:
.The parameter
GZ denotes the share of imported intermediate inputs in final-good production. Subject to
Φ3^� P�Z�� − �(1 − GZ)�H��*:,��: H��H + GZ �H�F*;,��: H��H � HH��Q, where ΦZ�� is real marginal cost of
9
downstream firm, likewise, downstream firm � minimizes the cost function L*:,�� �*:,��: +L*;,�� F*;,��:
to yield the following optimal demand for domestic and imported intermediates.
�*:,�� = (1 − GZ) RSIT,�4SI,�4 U�V �Z�� (12)
F*;,��: = GZ RSIW,�4SI,�4 U�V �Z�� (13)
The real marginal cost for downstream firm is derived as
ΦZ�� = L*�� = X(1 − GZ)�L*:,�� ��V + GZ�L*;,�� ��VY ���H (14)
L*�� is staggered producer price index for downstream firms. The final output is partly
consumed by domestic and foreign consumers with the remaining invested as capital stock to
produce upstream goods, �Z,�� ≡ _:,��: + �Z:,��; + #��.
2.2. Transportation cost
Following Ravn and Mazzenga (2004), we measure transportation cost as the wedge
between c.i.f. value of import and the corresponding f.o.b. value of export in such a way that
1 + `� = a�W,�4Tb�W,�4T = aIW,�4T
bIW,�4T = cW,�dTb[W,�4T (15)
2.3. Optimal pricing decision with U.S dollar pricing in trade
Pricing decision is assumed to be time dependent. The ability of domestic firms at
midstream and downstream production to re-optimize the price is subject to the signal
received at probability 1 − eSf, for g = 2,3. Firm � that receives the signal chooses ℙf:,� to
maximize the following expected discounted profits j�Π� for sales in home market
j�Π�:klD = j� ∑ (eSfn)�Λ�@� �ℙoT,�Ad4 �aco,�AdSo,�Ad � �ℙoT,�Ad4
SoT,�Ad��Mo,� �f:,�@��:∞�pq (16)
10
Contrary to producer-currency pricing decision in the typical New Keynesian model,
or local-currency pricing in the New Open-Economy model, we consider U.S. dollar pricing
strategy in exports. This assumption is apparently coherent with the fact that international
trade is largely denominated in the U.S dollar. The variability of exchange rates between
local currency and the U.S. dollar r:s,� will not pass through into foreign price of home
export, but rather, it will pass through into local-currency denominated export earnings. Firm
� signaled for price reoptimization chooses ℙf:,�s to maximize the expected export profit in
home currency
j�Π�tS = j� ∑ (eSf∗ n)�Λ�@� vwTx,�ℙoT,�Adx (�)�aco,�AdSo,�Ad y �wWx,�ℙoT,�Adx (�)
SoT,�AdW ��Mo,� �f:,�@��;∞�pq (17)
In what follows we assume that all firms are symmetric in price setting.
Firms allowed for price re-optimization will reset their log-linearized price ℙzf:,�fD{ to
approximate the optimal reset price derived from expected discounted profits (16) and (17),
respectively, for home and export market. The remaining firms that do not receive signal for
re-optimization will stick to last-period price, out of which a fraction of them |Sf will index
to last-period inflation. Probability-weighted inflation dynamics of producer price (}*:,�),
GDP deflator (}Z:,�), intermediate export price (}*:,�s ) and final export prices (}Z:,�s ) can be
derived as
}f:,� = + ~�o�@��o�~�o- }f:,��� + + ��@��o�~�o- j�}f:,�@� + ��Φz f,� + �f,� (18)
}f:,�s = 3 ~�ox�@��ox �~�ox 5 }f:,���s + 3 ��@��ox �~�ox 5 j�}f:,�@�s + �s�Φzf,� − �:s,� + �f,�s (19)
where � = (����o)(����o�)��o(�@��o�~�o) and �s = �����ox �����ox � ��ox ��@��ox �~�ox . �f,� is an i.i.d price markup shock for
g = 2,3. Φzf,� is the log-deviation of real marginal cost, in which Φz*,� = ��,� and ΦzZ,� = �*,�.
11
2.4. Household
Consider a continuum of infinitely-lived households, represented and indexed by
i∈ �0,1�, who possess the utility function of
� = j� �∑ n�u�c v(c�d���)������ − u�8 (8�d)�A�
�@� y∞�pq � (20)
where
_�� = v(|)���_:,��: ���� + (1 − |)���_;,��: ���� y ����
(21)
u�c and u�8 , respectively, arei.i.d preference and labor supply shock.
_:,��: �= 3 _:,��: (�)������ ���∈� 5 �������and_;,��: �= 3 _;,��: (�)������ ���∈� 5 ������� are the composite varieties
of home and imported consumer goods. ��(= �_���� ) indicates external habit formation in
which b is the parameter that governs the extent of habit persistence. 0 < n < 1 refers to
subjective discount factor, � measures the coefficient of relative risk aversion, and the
reciprocal of �indicates the wage elasticity of labor supply. The parameter � > 1 denotes the
elasticity of substitution between home and imported consumer goods. The parameter
| measures home bias. Household i’s constrained optimization problem can be illustrated as
maximization of utility function (20) subject to consumption basket (21) and the following
flow budget constraint
_� + +wTx,�S���- 3 ��W����5 + ��S��� + �� = 2� � + � + 01,����� + 3wTx,�����W @����S� 5 (22)
where L:,� and L;,� , respectively, denotes domestic price of home and imported consumer
goods. Household facing exchange-rate risk � in foreign asset market has access only to
imperfect international asset market. Note that foreign bond ¡�; is denominated in U.S. dollar.
Thus, the nominal exchange rate between home currency and the U.S. dollar is considered.
Solving for the utility maximization problem gives us the marginal rate of substitution
12
between works and consumption (23), intertemporal substitution of consumption (24), and
uncovered interest rate parity (25).
��� ��_�� − �_���� �u�8 = 2�a�w (23)
�c�d�¢c���d ��S� u�c = n(1 + 0�) �£�c�A�d �¢c�d ��
£�S�A� j�u�@�c (24)
r:s,� = j�r:s,�@� +�@<����@<� - � (25)
L� += ¤|L:,���¥ + (1 − |)L;,���¥¦� (��¥)⁄ - is the utility-based consumer price index (CPI).
Since household is a monopoly supplier of differentiated services, nominal wage is set
in Calvo-style, which results in nominal wage inflation dynamics as what follows:
}�{ = § ~¨�@�¨�~¨© }���{ + § ��@�¨�~¨© j�}�@�{ + �{(ª«�a�w − ª«� + u�{) (26)
where �{ = (���¨)(���¨�)�¨(�@�¨�~¨) . The parameter e{ denotes wage stickiness, and |{ is wage
indexation. u�? is i.i.d wage markup shock. Note that nominal wage inflation responds
positively to the wedge between wages demanded by optimizing household and marginal
product of labor.
2.5. Trade balance, aggregate value added, and monetary policy
We define trade balance as the balance between aggregate f.o.b exports and aggregate
c.i.f imports.
¬� = ��:,��; + �*:,��;�∈J �� + �Z:,��;�∈J �� − F�;,�: − F*;,��:�∈J �� − _;,��:�∈� �� (27)
The aggregate value added of the economy (), which corresponds to gross domestic product
in data,can be defined as
� = _� + #� + ¬� (28)
The model is lastly closed by considering a general form of monetary policy reaction as
below:
0� = ®�0��� + (1 − ®�)�0�f + °}cS�,� + z� + ∆w∆�:s,� + u�� (29)
13
where0�f+= u�c + �(²�� + ³�)- is the natural rate of interest influenced by the efficient shocks,
®� measures the interest rate persistence, °, , and ∆w , respectively, indicates central
bank’s responsiveness toward variability in CPI inflation, aggregate demand variability, and
rate of change in nominal exchange rates between home currency and U.S. dollar. u��refers to
i.i.d white noise to the conduct of monetary policy.
3. A Bayesian Measurement
The model is estimated using the Bayesian method. Bayesian estimation is principally
about identifying a set of parameters that maximize the posteriors µ(¶|¸, ℳ) as a product of
the likelihood function of data derived from the model µ(¸|¶, ℳ) using the Kalman filter
and the prior probability distribution of the parameters µ(¶, ℳ):
µ(¶|¸, ℳ) = µ(¸|¶,ℳ)µ(¶,ℳ) µ(¸|¶,ℳ)µ(¶,ℳ)s¶¶ (30)
where ¶ is a vector of model parameters, and ¸(= ℝ�, … , ℝ¼) is a set of observed data over
½ number of periods (see, for instance, Fernandez-Villaverde, 2010 for detailed discussion).
3.1. Data
In addition to China, we study nine East and Southeast Asian economies including
Japan, the Republic of Korea, Hong Kong, Taiwan, Singapore, Malaysia, Thailand, Indonesia,
and the Philippines. The country selection is motivated by the fact that these Asian
economies are vertically specialized in trade. Whereas production sharing is equally observed
in the rest of the world, the depth and complexity of production fragmentation and trade in
East Asia is unparalleled. In a study of 79 countries and over 121 categories of goods from
1967-2005, Amador and Cabral (2009) show that out of the top ten vertically specialized
countries, eight are located in East Asia (see, also, Sawyer et al., 2010). Furthermore,
14
Koopman et al. (2010) argue that the positions of advanced and emerging Asian countries in
the global production chain of the manufacturing sector are diverse with Japan occupying the
most upstream production, Korea and Taiwan in midstream production, and China in the
most downstream production. These factors make Asian economies an ideal sample for our
investigation.
We analyze the quarterly time series from 2001 to 2008 due to the fact that the
addition of China to the World Trade Organization at the end of 2001 fundamentally changed
regional production sharing by substituting Southeast Asia as the destination for final
assembly of intermediates into consumer goods exported to the United States, Europe, and
the rest of the world (see, for instance, Kim et al. 2011, and Athukorala, 2009). Excluding
China (CN), we categorize Indonesia, Malaysia, Philippines, Thailand as developing
Southeast Asian economies (SEA4) and Japan, the Republic of Korea, Hong Kong, Taiwan,
and Singapore, as advanced East Asian economies (EA5). Thus, there are three two-country
models to be estimated: the SEA4-EA5 model, the CN-EA5 model, and the CN-SEA4 model.
The name that appears first is treated as home country and the name that appears second is
treated as a foreign country.
There are nineteen macroeconomic observable variables used in estimation, including
GDP; consumption; investment; labor force; nominal short-term interest rate; nominal
exchange rates between home currency and the U.S. dollar; PPI inflation; GDP deflator
inflation; CPI inflation for SEA4, EA5, and China in a two-country setting; and the U.S.
federal funds rate. All the quantity variables are deflated by the respective deflators, and all
data, except for the rates of inflation and interest, are logged and de-trended using the
Hodrik-Prescott filter with a smoothing parameter of � = 1600. We then construct a cyclical
observable series for SEA4 and EA5 weighted by the time-varying fraction of national total
trade over aggregate regional trades.
15
3.2. Priors and posteriors
Table 1 reports the priors and probability distribution functions. We assume symmetric
priors for both home and foreign country. Nonetheless, we allow different posteriors for price
indexation and stickiness, the share of imported intermediate inputs, home bias in
consumption, monetary policy reaction functions, and standard deviation of structural shocks.
We use Dynare 4.3 algorithms for model estimation and adjust the number of Markov chains
to ensure that estimates for mean and standard deviation within and across Markov chains are
consistent.
In addition, we have assumed a uniform distribution for priors where the true value is
uncertain because of the lack of previous Bayesian studies on this region. The priors include
the share of imported inputs at midstream and downstream production, price stickiness, and
the standard deviation of shocks. Thus, equal probability is assigned for all potential
parameter values within a theoretically coherent range. The priors for efficient shock
persistence are in beta distribution, while the parameters in monetary policy reaction function,
which theoretically must have positive values, are in gamma distribution with prior means
following the standard assumption.
[INSERT TABLE 1 HERE]
Table 2 reports the estimates of mode, mean and probability interval for selected
parameters in the SEA4-EA5 model, the CN-EA5 model, and the CN-SEA4 model. All
structural shocks and parameters are nicely identified, as evidenced by the proximity between
posterior mode and mean, which falls within the 90% probability interval. Two intermediate
input parameter estimates are particularly worthy of greater discussion. First, Table 2 shows
that China, Southeast Asia, and East Asia are tightly bound in the sense that midstream and
downstream productions in these economies heavily use each other’s export of intermediate
inputs.
16
Second, the estimated elasticity of substitution between domestic and imported
intermediates is greater than one, irrespective of models. This finding is in interesting
contrast to the existing empirical finding of inelastic substitution because of the
complementarities between domestic and foreign inputs (see, for instance, Luong, 2011).
However, such empirical estimates are about the elasticity of substitution between
intermediate inputs across different final outputs, whereas elasticity of substitution between
intermediates is within a final output sector in our context. More importantly, di Giovanni
and Levchenko (2010) argued that the elasticity of substitution is not as empirically important
in the bilateral business cycle comovement as the literature on trade-comovement puzzle has
theoretically speculated (see, for instance, Kose and Yi, 2006).
[INSERT TABLE 2 HERE]
4. Tracing value added, production fragmentation, and country upstreamness
Koopman et al. (2010) offer the most detailed definition of vertical specialization to
date by breaking down gross exports into the foreign value added in domestic exports of
intermediates and final goods – the earliest definition of vertical specialization by Hummels
et al. (2001) – and the domestic value added embodied in the exports of the following:
(1) final goods and services absorbed by the direct importer;
(2) intermediates used by the direct importer to produce its domestically required
products;
(3) intermediates used by the direct importer to produce goods for third countries; and
(4) intermediates used by the direct importer to produce goods shipped back to source.
4.1. Decomposing the value added of gross exports
We contribute to this literature by deriving an equation from the estimated model that
corresponds exactly in definition to the Koopman et al. (2010) decomposition. This exercise
is useful in two mutually beneficial ways. First, it provides a macroeconomic perspective of
17
vertical specialization that corresponds to the micro evidence, paving the way for the
computation of upstreamness of countries engaged in global production sharing vis-à-vis
trading partner. Second, the use of the international business cycle model to explain the
macroeconomic implications of vertical specialization is more transparent because of its
ability to account for the different types of value added in trade.
To be compatible with the decomposition of Koopman et al. (2010), we break down
gross exports to domestic value added embodied in exports ¿¯", “reflected domestic value
added” ¯r1 ∗, and the foreign value added used in exports, ¯r:
À� = Á ¿¯"f,�f + ¯r1 ∗�+ ¯r�
where
∑ ¿¯"f,�f =)1 − aIW,�T
b[� R1 − a�T,�WbI�W U/ �Z:,�;ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(�)ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ�
+ 31 − aÆW,�Tb�� 5 Ra�T,�W
bI�W U �*;,�;ÂÃÃÃÃÃÃÄÃÃÃÃÃÃÅ(��)+ 31 − a�W,�T
bI� 5 RaIT,�Wb[�W U �_;,�; + #�; ÂÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÅ(���)ÂÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÅ*
(31)
¯r1 ∗�= 31 − aÆW,�Tb�� 5 Ra�T,�W
bI�W U �*;,�:ÂÃÃÃÃÃÃÄÃÃÃÃÃÃÅ(�Ç)+ 31 − a�W,�T
bI� 5 RaIT,�Wb[�W U �Z;,�:ÂÃÃÃÃÃÃÄÃÃÃÃÃÃÅ(Ç)ÂÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÅÈ
(32)
¯r� = R1 − aÆT,�Wb��W U 3a�W,�T
bI� 5 �*:,�;ÂÃÃÃÃÃÃÄÃÃÃÃÃÃÅ(Ç�)+ R1 − a�T,�W
bI�W U 3aIW,�Tb[� 5 �Z:,�;ÂÃÃÃÃÃÃÄÃÃÃÃÃÃÅ(Ç��)
(33)
By adding and simplifying equations (31) to (33) together with the transportation cost, we
obtain the standard definition of gross exports, À� ≡ ∑ �f:,�;Zfp� . Thus, it verifies our
decomposition.
Component (1) in equation (31) represents the domestic value added from the export
of final goods and is calculated using the residuals between final output and the imported
intermediates in gross output. Because imported intermediates F*;,�: comprise domestic value
18
added embodied in an earlier stage of foreign production, we add F�:,�; as a share of gross
foreign midstream output back into component (1). Component (2) in equation (31) refers to
the domestic value added embodied in the export of domestic intermediates for foreign local
needs at midstream F�:,�; (item (ii)) and downstream level F*:,�;
(item (iii)) as a share of
gross foreign output, respectively. Domestic intermediate production contains foreign value
added because foreign intermediates from an earlier stage were used. This value must be
removed. However, Fq;,�: = 0 because upstream production only uses domestically owned
inputs.
In contrast to component (2), component (4) in equation (32) indicates domestic value
added embodied in the domestic intermediates used in foreign production for re-exporting
back to the source as intermediates (item (iv)) or final goods (item (v)). Likewise, foreign
value added incorporated at an earlier stage of production of domestic intermediates must be
taken into account. Finally, equation (33) informs us about the foreign content of domestic
exports, after accounting for the domestic value added embodied in foreign intermediates.
Consider the definition of the log-linearized variable: É� = ÊËÌ +Í�ÍÎ -, where ÏÎ refers to
the steady-state value. It implies that Ï� = ÏÎ� Ð�. By considering the relationship between c.i.f
import and f.o.b export, as in equation (15), our decomposition of gross export can be
rewritten and rearranged for measurement as follows:
1 = )1 − GZ R1 − ÑI��@ÒÎW DÓ�WU/ 3ÔÎ[TW
ÀÕ 5 �+Ô[T,�W �À«�-ÂÃÃÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÃÃÅ(�)+ ) ÑIW��@ÒÎW DÓ�W/ RÔÎIWW
ÀÕ U �+ÔIW,�W �À«�-ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(��)+
(1 − G*) ) Ñ[W��@ÒÎW DÓ�W/ )RÖWWÀÕ U �+ÖW,�W �À«�- − +ØWÀÕ - �+Ø�W�À«�-/ÂÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÃÃÃÃÃÅ(���)
+ ) ÑIW��@ÒÎW DÓ�W/ 3ÔÎIWTÀÕ 5 �+ÔIW,�T �À«�-ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(�Ç)
+
19
(1 − G*) ) Ñ[W��@ÒÎW DÓ�W/ 3ÔÎ[WTÀÕ 5 �+Ô[W,�T �À«�-ÂÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÅ(Ç)
+ G* 3ÔÎITWÀÕ 5 �+ÔIT,�W �À«�-ÂÃÃÃÃÃÄÃÃÃÃÃÅ(Ç�)
+
)1 − ÑIW��@ÒÎW DÓ�W/ GZ 3ÔÎ[TWÀÕ 5 �+Ô[T,�W �À«�-ÂÃÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÃÅ(Ç��)
(34)
where Gf ≡ Ff��,;,�: �f,�$ .
4.2. Relation to existing models and measurements
How important is it to set three sequential stages of production where outputs of all of
the chains can be traded? To answer this question, we consider three different dominant
models of international business cycles with trade: the international real business cycle model
of Raffo (2008), the open-economy New Keynesian model of Smets and Wouters (2003), and
the dynamic Ricardian trade model of Yi (2003).
Raffo (2008) considers two-stage productions where domestic intermediates cross
international borders for foreign final production, and vice versa, while final goods are non-
traded. Smets and Wouters (2003) also employ a two-stage production structure but allow
both intermediates and final goods to be traded. Yi (2003) resembles our model the most, in
that the final good is produced in three sequential stages. The notable difference is that final
goods are not tradable in Yi (2003), although both upstream and midstream outputs are. By
matching equations (31) to (33) to the trade structure of each of the aforementioned models,
the value-added decomposition of gross exports in these models can be written as:
International RBC of Raffo (2008): À� = 31 − a�W,�TbI� 5 RaIT,�W
b[�W U �_;,�; + #�; ÂÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÅ(���)
NK model of Smets and Wouters (2003):
20
À� = �1 − F*;,�:�Z� Ù1 − F�:,�;
�*�; Ú� �Z:,�;ÂÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÅ(�)+ R1 − F�;,�:
�*� U ÙF*:,�;�Z�; Ú �_;,�; + #�; ÂÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÅ(���)
+ R1 − F�;,�:�*� U ÙF*:,�;
�Z�; Ú �Z;,�:ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(Ç)+ Ù1 − F�:,�;
�*�; Ú RF*;,�:�Z� U �Z:,�;ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(Ç��)
Dynamic Ricardian trade model of Yi (2003):
À� = R1 − Fq;,�:��� U ÙF�:,�;
�*�; Ú �*;,�;ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(��)+ R1 − F�;,�:
�*� U ÙF*:,�;�Z�; Ú �_;,�; + #�; ÂÃÃÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÃÃÅ(���)
+ R1 − Fq;,�:��� U ÙF�:,�;
�*�; Ú �*;,�:ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(�Ç)+ Ù1 − Fq:,�;
���; Ú RF�;,�:�*� U �*:,�;ÂÃÃÃÃÃÃÃÄÃÃÃÃÃÃÃÅ(Ç�)
Many important genres of value added in domestic and foreign trade have been left out
of these models. The case of Raffo (2008) supported the argument of Hummels et al. (2001)
that vertical specialization is more than trade in intermediate goods. Vertical specialization
involves exports and imports. Beside, as shown in Table 3, the missing domestic value added
in intermediates exported back to the source in Smets and Wouters (2003) and the foreign
value added in domestic exports Yi (2003) are too critical to be ignored when measuring the
country upstreamness in production sharing.
Constrained by a two-country setting, our model is unable to account for the domestic
value added embodied in the export of intermediates that will be re-exported to a third
country by a direct importer after processing. Even so, relative to the existing models, our
model permits the most comprehensive break down by matching the Koopman et al. (2010)
decompositions so that, divided by the gross export, ∑ ¿¯"f,�f is the VAX ratio of Johnson
and Noguera (2012a, 2012b); ¯r1 ∗� corresponds partly to VS1 by Hummels et al. (2001)
and exactly to VS1* in Daudin et al. (2011); and ¯r� matches the Hummels et al. (2001) ¯r
definition.
21
4.3. Identifying production fragmentation and the upstreamness of countries
Based on the estimated Gf , the calibrated “great ratios” for the market-clearing
condition of each stage, the steady-state transportation cost, and the data generated from the
estimated model, Table 3 provides a numerical simulation of the models.
[INSERT TABLE 3 HERE]
Looking first at bilateral vertical specialization in gross exports (VS), the estimates
range on average from 20.11% to 34.97% during the period from 2001Q1 to 2008Q4.
Although not exactly comparable, this measure is aligned with the estimates of multilateral
vertical specialization of gross export obtained by Koopman et al. (2010) that range from
23.65% for EA5 (excluding Singapore), to 32.28% for SEA4 and to 33.8% for China in 2004
for the manufacturing sector. In addition, the measures are comparable to those obtained in
Amador and Cabral (2009) that, as a share of gross import, range from 20.3% for China to
24.5% for EA5 (excluding Japan) and to 28.9% for SEA4 (excluding Indonesia) during the
period from 2001-2005.
The VS ratio is imprecise as a measure of production fragmentation. For example,
suppose Country A produces only at the downstream stage for final goods export with
intermediates imported entirely from abroad. By contrast, Country B slices up both the
midstream and downstream production with approximately half of the intermediates imported.
The vertical specialization as a share of gross exports for Country A is equivalent to Country
B despite the fact that the latter has more fragmented production chains. The VAX ratio
proposed by Johnson and Noguera (2012a, 2012b) is more intuitively straightforward as a
measure of production fragmentation.
If the exports do not contain imported intermediates, which implies the absence of
production fragmentation, then the VAX ratio will be 100%. In other words, as Johnson and
Noguera (2012b) argued, the ratio declines when foreign intermediates are used in the final
22
goods export to the direct importer ((i) in Table 3) and when exported intermediates are used
for local needs ((ii) and (iii)). Back-and-forth trade in intermediates, made possible by
production fragmentation, also multiplies gross exports more proportionally than value added,
depressing the ratio further.
That said, by using the bilateral VAX ratio, Table 3 clearly shows that production in
advanced East Asian economies has been on average more vertically fragmented than in
Southeast Asian economies (22.85% versus 72.54%) and China (25% versus 59.13%). In the
case of China and Southeast Asia, China is slightly ahead of Southeast Asia (43.08% versus
52.29%).
Decomposing gross exports into domestic and foreign value added facilitates the
construction of an index that helps us to trace the upstreamness of a country within the global
production sharing chain. Intuitively, for countries engaged upstream of the production
sharing, intermediate exports used by a foreign importer are more likely to be shipped back
home for further processing. As a result, reflected domestic value added in the gross exports
of upstream countries will be more than proportionate to the foreign value added in its gross
exports.
Thus, in the spirit of Koopman et al. (2010), we propose an upstreamness index
computed as the home country’s log ratio of ¯r1 ∗ to ¯r in gross exports.
� = Êg +1 + Ûw�∗À� - − Êg +1 + ÛwÀ� - (35)
Not surprisingly, Table 3 shows that East Asian economies with an index of more than one
stay at the upper end of the production chain compared to Southeast Asia and China. This
largely confirms the findings of Kim et al. (2011), Koopman et al. (2010) and Athukorala
(2009): Japan lies upstream in the production chain by providing intermediate inputs and
China lies downstream in the production chain by using imported intermediate inputs to
23
provide final goods to the world, while other Asian economies, including Hong Kong, Korea,
Taiwan, Malaysia, and Philippines, lie in between.
What makes our story interesting is that both China and Southeast Asia lie
downstream in their bilateral interaction. This suggests that while China complements
advanced East Asian economies as an absorber of the latter’s capital and intermediate goods,
China competes head-to-head with Southeast Asia at the final goods market (see, for instance,
Haltmaier et al., 2007 and Eichengreen et al., 2007).
5. Does country upstreamness matter for international shock transmission?
When foreign intermediates are heavily used and productions are highly fragmented
and linked, as Table 3 has shown, one expects greater synchronization in the business cycles
of these economies. In fact, di Giovanni and Lechvenko (2010) found that vertical trade is too
important of a determinant to bilateral business cycle comovement to be ignored. He and
Liao (2012), Moneta and Ruffer (2009), and Shin and Wang (2003), to name few, have found
increasingly synchronized business cycles in Asia made possible by intra-industry trade.
Figure 2 vividly illustrates this point by comparing bilateral cross-correlation of simulated
data with the actual series across three estimated models. During the period of 2001 to 2008,
these economies comove tightly wherein most bilateral exports and imports are synchronized
contemporaneously, followed by gross domestic product and consumption. And our
estimated model of production fragmentation and vertical trade is able to replicate the
business cycle comovement, although not perfectly.
[INSERT FIGURE 2 HERE]
Business cycle synchronization, made possible by trade integration, is often viewed as
the underlying condition for the formation of a monetary union. However, business cycle
synchronization does not necessarily imply symmetric responses toward macroeconomic
24
shocks. While the business cycles of two countries that are heavily vertically linked in
production and trade comove over a period in response to a variety of shocks, they may
respond asymmetrically towards an identical shock. In fact, the seminal Bayoumi and
Eichengreen (1994) have long noted that “output comovements are not the same as shocks.”
Even studies attempting to distinguish common from country-specific shocks confuse
disturbances and responses. We believe that more light needs to be shed on the latter in the
discussion of macroeconomic interdependence.
5.1. Supply shock
Consider a 1% increase in the innovation of total factor productivity (TFP). We assess
the dynamic response of the aggregate value added (GDP) of each region towards both
common – if home and foreign TFP shock are perfectly correlated – and regional-specific
TFP shock, as shown in Figure 3. Vertical reading of the region indicates the place where
shock originates, whereas horizontal reading refers to the responses. As far as idiosyncratic
shock is concerned, China’s favorable TFP shock marginally increases the GDP of the East
Asian countries engaged in upstream production, but not of Southeast Asia, which is involved
downstream in bilateral production fragmentation.
[INSERT FIGURE 3 HERE]
By contrast, positive shocks to downstream Southeast Asian TFP that expand GDP
help downstream China’s GDP to marginally prosper; however, it does not affect upstream
East Asian economies. It is puzzling that favorable shock to upstream East Asian TFP
expands Chinese GDP but not Southeast Asian GDP, even though both are positioned
downstream from East Asia in the production sharing chain. This pattern is robust even when
the TFP shocks to China and Southeast Asia are perfectly correlated with the East Asian TFP
shock. It should be noted that when the TFP shock originates in downstream China, where
the other countries are perfectly correlated, the effect is benign to all regardless of their
25
position in the production sharing chain. What is puzzling is that this is not the case when the
TFP shock originates in downstream Southeast Asia.
Therefore, based on the case of China-East Asia, we can cautiously infer that when an
efficient supply shock strikes, the responses of the countries positioned at different stages of
production that complement one another tend to be synchronized. Nonetheless, for countries
engaged in the downstream production phase, expansion may (or may not) negatively affect
production partner countries.
5.2. Demand shock
However, when facing a demand shock, responses are identical, regardless of the
position in the production sharing chain. Figure 4 illustrates the dynamic response of GDP
when there is a 1% negative shock to the innovation of preference. As a consequence, GDP
of all regions declines. Nevertheless, production upstreamness leaves a mark on the pattern of
responses. When shocks originate in downstream China or Southeast Asia, to which the East
Asian preference shock is correlated, the response of upstream East Asia becomes larger. By
contrast, if the shock originates in the upstream East Asian preference, the responses of
downstream China and Southeast Asia are larger only if the shock is idiosyncratic. For
countries that stay equally in the downstream, the pattern of response is not conclusive. In
particular, Chinese GDP falls more if the shock to the Southeast Asian preference is
idiosyncratic. Quite the opposite, the magnitude of contraction of Southeast Asian GDP is
stronger if the shock to the Chinese preference is correlated with its own.
[INSERT FIGURE 4 HERE]
5.3. Relation to the empirical literature on the great trade collapse and vertical
specialization
Our results indicate that the role of vertical specialization and production upstreamness
in shock propagation, especially in the case of East Asia and China when facing a negative
26
preference shock, corroborates the existing findings on vertical specialization and the recent
great trade collapse. For instance, Levchenko et al. (2010) illustrated that downstream
vertical linkages play a role in the reduction of international trade, which, in turn, contributes
to the collapse of world aggregate value added. In particular, goods that are intensively used
as intermediates for other countries experienced larger percentage drops in imports and
exports. Framed in our context, in the face of negative demand shock the upstream countries
tend to be affected more severely than the downstream countries.
Bems et al. (2010) maintained that both intermediate and final exports from Japan fall
more than exports from China, despite the fact that the United States is a larger direct
importer for China than Japan. Put in our context, the ripples of negative demand shock that
originated in an extra-regional country must first arrive at downstream China and, to some
extent, upstream East Asia, which also exports final goods in production sharing. “Third-
country shock” is akin to common shock in our model with its origins in China. The pattern
of responses of the GDP of China and East Asia towards negative demand shock shown in
Figure 4 fits with this evidence.
The intuition is straightforward. Common shock hits East Asia harder2
because
upstream East Asia is more intensely involved in production sharing with multiple back-and-
forth import and export of intermediates. This is evidenced by higher domestic value added
embodied in intermediates exported back to East Asia for further processing at the midstream
level vis-à-vis China (30.73% versus 7.86%) and Southeast Asia (39.05% versus 0.37%) in
Table 4.
2 Wang and Whalley (2010) document the export performance of Japan, the Republic of Korea, Taiwan,
Indonesia, Malaysia, Thailand, and China over the periods from May 2008 to December 2009, with which our
story depicted in Figure 3 under common shock is compatible. For instance, in the case of the United State as
direct importer, on simple average, month-to-month export performance of the advanced East Asian economies
deteriorates the most by 20.98%, followed by the developing Southeast Asia at 12.64% and China at 3.7%.
27
6. Conclusion
This paper adds to the literature on the measurement of production fragmentation and
value added in trade. However, contrary to the admirably microscopic approach in empirical
research, this paper takes up the challenge through a Bayesian estimated New Keynesian
model expanded to feature three sequential production stages. Outputs of all stages are
tradable. The estimated gross exports from this expanded New Keynesian model can then be
broken down into domestic and foreign value added that are compatible with the most
comprehensive classifications of value added in trade available in the empirical literature.
The usefulness of this macroeconomic approach is twofold. One the one hand, it makes
a model that attempts to explain the macroeconomic implications of vertical specialization
transparent in terms of its ability to account for the types of value added in trade; therefore, it
provides a yardstick on the empirical relevance of the model. On the other hand, it is easier to
proceed to a follow-up investigation on the macroeconomic implications of vertical
specialization in a theoretically and empirically coherent manner. For instance, we find that
the responses of countries bound but positioned at different stages in production sharing tend
to be synchronized even in the face of a country-specific supply shock. In addition, the
magnitude of the response towards common and country-specific shocks is shaped by the
country upstreamness. Upstream countries tend to be more responsive when the shock is
common and originates in downstream countries.
A limitation of the two-country model is its inability to account for the domestic value
added embodied in the export of intermediates that will be re-exported to a third country by a
direct importer after processing. The role of the third country can be important in the trade
dynamics of production sharing. The United States and Europe, for instance, are heavyweight
extra-regional trading partners for East Asia. Setting up and estimating a multi-country model
shall be the venue for the next pursuit.
28
References
Amador, J., Cabral, S., 2009. Vertical specialization across the world: a relative measure. The
North American Journal of Economics and Finance 20(3), 267-280.
Athukorala, P-C., 2009.The rise of China and East Asian export performance: is the
crowding-out fear warranted? The World Economy 32(2), 234-266.
Bayoumi, T., Eichengreen, B., 1994. One money or many? Analyzing the prospects for
monetary unification in various parts of the world. Princeton Studies in International
Finance No. 76, Princeton, New Jersey.
Bems, R., Johnson, R.C., Yi, K.-M., 2010. Demand spillovers and the collapse of trade in the
global recession. IMF Economic Review 58(2), 295-326.
di Giovanni, J., Levchenko, A.A., 2010. Putting the parts together: Trade, vertical linkages,
and business cycle comovement. American Economic Journal: Macroeconomics 2(2), 95-
124.
Daudin, G., Rifflart, C., Schweisguth, D., 2011. Who produces for whom in the world
economy? Canadian Journal of Economics 44(4), 1403-1437.
Eichengreen, B., Rhee, Y., Tong, H., 2007. China and the exports of other Asian countries.
Review of World Economics 143(2), 201-226.
Feenstra, R., 1998. Integration of trade and disintegration of production in the global
economy. Journal of Economic Perspectives 12(4), 31-50.
Fernandez-Villaverde, J., 2010. The econometrics of DSGE models. SERIES: Journal of the
Spanish Economic Association 1, 3-49.
Goldberg, L., Tille, C., 2008. Vehicle currency use in international trade. Journal of
International Economics 76(2), 177-192.
Haltmaier, J.T., Ahmed, S., Coulibaly, B., Knippenberg, R., 2007. The role of China in Asia:
engine, conduit or streamroller? International Finance Discussion Papers No. 904, Board
of Governors of the Federal Reserve System.
He, D., Liao, W., 2012. Asian business cycle synchronization. Pacific Economic Review
17(1), 106-135.
Hummels, D., Ishii, J., Yi, K-M., 2001. The nature and growth of vertical specialization in
world trade. Journal of International Economics 54, 75-96.
Johnson, R.C., Noguera, G., 2012a. Accounting for intermediates: Production sharing and
trade in value added. Journal of International Economics 86(2), 224-236.
Johnson, R.C., Noguera, G., 2012b. Fragmentation and trade in value added over four
decades. Unpublished manuscript, Dartmouth College.
Kim, S., Lee, J-W, Park, C-Y., 2011. Ties binding Asia, Europe and the USA. China &
World Economy 19(1), 24-46.
Koopman, R., Powers, W., Wang, Z., Wei, S.-J., 2010. Give credit where credit is due:
tracing value added in global production chains. NBER Working Paper No. 16426. NBER,
Cambridge.
Levchenko, A.A., Lewis, L.T., Tesar, L.L., 2010. The collapse of international trade during
the 2008-09 crisis: In search of the smoking gun. IMF Economic Review 58(2), 214-253.
Luong, T.A., 2011. The impact of input and output tariffs on firms’ productivity: Theory and
evidence. Review of International Economics 19(5), 821-835.
Mandelman, F.S., Rabanal P., Rubio-Ramirez, J. F., Villan, D., 2011. Investment-specific
technology shocks and international business cycles: An empirical assessment. Review of
Economic Dynamics 14(1), 136-155.
Moneta, F., Ruffer, R., 2009. Business cycle synchronisation in East Asia. Journal of Asian
Economics 20, 1-12.
29
Raffo, A., 2008. Net exports, consumption volatility and international business cycle model.
Journal of International Economics 75, 14-29.
Ravn, M.O., Mazzenga, E., 2004. International business cycles: the quantitative role of
transportation costs. Journal of International Money and Finance 23(4), 645-671.
Sawyer, W.C., Sprinkle, R.L., Tochkov, K., 2010. Patterns and determinants of intra-industry
trade in Asia. Journal of Asian Economics 21(5), 485-493.
Shin, K., Wang, Y., 2003. Trade integration and business cycle synchronization in East Asia.
Asian Economic Papers 2(3), 1-20.
Smets, F., Wouters, R., 2003. An estimated dynamic stochastic general equilibrium model of
the Euro Area. Journal of the European Economic Association 1(5), 1123-1175.
Wang, J., Whalley, J., 2010. The trade performance of Asian Economies during and
following the 2008 financial crisis. NBER Working Paper No. 16142, National Bureau of
Economic Research.
Yi, K.-M., 2003. Can vertical specialization explain the growth of world trade? Journal of
Political Economy 111(1), 52-102.
30
Table 1 The priors for parameters and shocks
Prior distribution
Probability
distribution
function
Mean Standard
deviation
Parameters
Risk aversion coefficient � Uniform 1 0.577
Reciprocal of wage elasticity of labor supply �8 Gamma 2 1.000
Habit persistence � Beta 0.7 0.100
Forward looking-ness of investment decision Λ Uniform 0.5 0.289
Els btw. home and imported intermediate goods O Normal 1.5 0.500
Home bias in consumption | Beta 0.7 0.100
Share of imported materials at intermediate production G* Uniform 0.5 0.289
Share of imported intermediate goods at final production GZ Uniform 0.5 0.289
Employment indexation |{ Uniform 0.5 0.289
Producer price indexation |Ü* Uniform 0.5 0.289
Final goods price indexation |ÜZ Uniform 0.5 0.289
Intermediate export price indexation |Ü*∗ Uniform 0.5 0.289
Final goods export price indexation |ÜZ∗ Uniform 0.5 0.289
Employment stickiness eD Uniform 0.75 0.144
Producer price stickiness eÜ* Uniform 0.75 0.144
Final goods price stickiness eÜZ Uniform 0.75 0.144
Intermediate export price stickiness eÜ*∗ Uniform 0.75 0.144
Final export price stickiness eÜZ∗ Uniform 0.75 0.144
Policy inertia ®� Beta 0.7 0.100
Policy response to inflation ° Gamma 1.5 1.000
Policy response to GDP fluctuation Gamma 0.125 0.050
Policy response to exchange rate variability ∆w Gamma 0.5 0.100
TFP shock persistence ®Ý Beta 0.8 0.100
IST shock persistence ®� Beta 0.7 0.100
Shocks
Total factor productivity �Ý Uniform 0.5 0.289
Investment-specific technology �� Uniform 0.5 0.289
Labor supply �f Uniform 0.5 0.289
Preference �Ö Uniform 0.5 0.289
Producer price markup �°IT Uniform 0.5 0.289
Final goods price markup �°[T Uniform 0.5 0.289
Intermediate export price markup �°IT∗ Uniform 0.5 0.289
Final export price markup �°[T∗ Uniform 0.5 0.289
Transportation cost �Ò Uniform 0.5 0.289
Monetary policy �< Uniform 0.5 0.289
UIPC �Þ Uniform 0.5 0.289
U.S monetary policy �;;< Uniform 0.5 0.289
31
Table 2 Selected posterior distributions for the estimated two-region models, 2001Q1-2008Q4
Southeast Asia (SEA4) East Asia (EA5)
Mode 5% Mean 95% Mode 5% Mean 95%
Parameters
� 0.293 0.251 0.307 0.367 0.293 0.251 0.307 0.367 | 0.636 0.559 0.618 0.669 0.891 0.858 0.896 0.928 O 1.374 1.354 1.402 1.447 1.374 1.354 1.402 1.447
Λ 0.954 0.919 0.963 1.000 0.954 0.919 0.963 1.000 G* 0.434 0.238 0.378 0.507 0.724 0.629 0.797 0.955 GZ 0.199 0.110 0.197 0.277 0.579 0.415 0.524 0.628 eÜ* 0.751 0.708 0.732 0.754 0.724 0.649 0.688 0.728 eÜZ 0.883 0.868 0.883 0.899 0.889 0.876 0.891 0.906 eÜ*∗ 0.907 0.864 0.895 0.926 0.505 0.500 0.529 0.558 eÜZ∗ 0.712 0.650 0.693 0.732 0.726 0.671 0.712 0.764
China (CN) East Asia (EA5)
Mode 5% Mean 95% Mode 5% Mean 95%
� 0.601 0.499 0.637 0.787 0.601 0.499 0.637 0.787 | 0.479 0.443 0.480 0.517 0.788 0.754 0.791 0.833 O 1.476 1.468 1.490 1.514 1.476 1.468 1.490 1.514
Λ 0.922 0.846 0.905 0.971 0.922 0.846 0.905 0.971 G* 0.405 0.339 0.411 0.497 0.493 0.450 0.548 0.640 GZ 0.999 0.932 0.965 1.000 0.733 0.583 0.693 0.812 eÜ* 0.849 0.824 0.842 0.859 0.797 0.781 0.802 0.825 eÜZ 0.931 0.902 0.922 0.940 0.925 0.914 0.923 0.931 eÜ*∗ 0.715 0.644 0.708 0.761 0.746 0.726 0.783 0.837 eÜZ∗ 0.729 0.705 0.733 0.759 0.784 0.761 0.781 0.805
China (CN) Southeast Asia (SEA4)
Mode 5% Mean 95% Mode 5% Mean 95%
� 0.916 0.812 1.020 1.209 0.916 0.812 1.020 1.209 | 0.844 0.808 0.862 0.902 0.603 0.579 0.618 0.663 O 1.592 1.519 1.559 1.603 1.592 1.519 1.559 1.603
Λ 0.927 0.861 0.920 0.991 0.927 0.861 0.920 0.991 G* 0.512 0.377 0.518 0.709 0.732 0.424 0.610 0.771 GZ 1.000 0.901 0.949 1.000 0.637 0.597 0.752 0.919 eÜ* 0.656 0.623 0.657 0.694 0.567 0.513 0.563 0.611 eÜZ 0.935 0.927 0.941 0.954 0.856 0.835 0.855 0.874 eÜ*∗ 0.745 0.664 0.744 0.813 0.648 0.523 0.643 0.732 eÜZ∗ 0.715 0.677 0.707 0.739 0.555 0.500 0.522 0.546
Notes: (i) The posterior distribution is obtained using the Metropolis-Hastings sampling algorithm based on 4
parallel chains of 50,000 draws, of which the first half was discarded as burn-in. The average acceptance rate for
each estimated model is as what follows: SEA4-EA5: 0.219; CN-EA5: 0.235; and CN-SEA4: 0.251. We impose
identical posteriors for �, O, and Λ across regions. (ii) SEA4 comprises Indonesia, Malaysia, the Philippines,
and Thailand weighted by total trades. (iii) EA5 comprises Japan, Hong Kong, Korea, Singapore, and Taiwan
weighted by total trades.
32
Table 3 Measuring average value added in gross export, and production fragmentation and upstreamness
Decomposition of value added in gross exports (%)
Connection with existing
measures
Measure of
Upstreamness
DVA in
direct
final
goods
export
DVA in intermediates
absorbed by foreign for
domestic use
DVA in intermediates export
shipped back to source Foreign VA
VAX VS1* VS �
Midstream Downstream Midstream Downstream Midstream Downstream
(i) (ii) (iii) (iv) (v) (vi) (vii)
SEA4-EA5
Southeast Asia 7.374 35.793 29.369 0.373 6.983 19.733 0.376 72.536 7.355 20.109 0.366
East Asia 17.633 4.132 1.089 39.053 3.126 26.052 8.916 22.853 42.179 34.968 1.206
CN-EA5
China 23.723 19.164 16.237 7.856 0.422 12.762 19.835 59.125 8.278 32.597 0.254
East Asia 11.426 7.761 5.754 30.731 11.781 23.656 8.891 24.941 42.512 32.547 1.306
CN-SEA4
China 35.839 4.323 2.917 9.307 10.580 11.907 25.125 43.080 19.888 37.032 0.537
Southeast Asia 28.751 15.918 7.624 11.041 1.876 15.433 19.358 52.293 12.917 34.790 0.371
Notes: (i) As a share of gross exports, VAX refers to domestic value added proposed by Johnson and Noguera (2012a, 2012b), and is a sum of (i), (ii), and (iii); VS1* is
reflected domestic value added, defined in Daudin et al. (2011) and responds partly to Hummels et al. (2001), and equals (iv) + (v); VS refers to foreign value added of
domestic exports as defined by Hummels et al. (2001), and is the sum of (vi) and (vii).
(ii) � = Ûw�∗Ûw .
(iii) SEA4 comprises Indonesia, Malaysia, the Philippines, and Thailand weighted by total trades.
(iv) EA5 comprises Japan, Hong Kong, Korea, Singapore, and Taiwan weighted by total trades.
33
Home
Foreign
Technology: �� = ß(, �)
Market: �� ≡ ��: + ��:;
��; = �; , �;
��; ≡ ��;: + ��;;
Technology: �* = ���:, F�;:
Market: �* ≡ �*: + �*:;
�*; = ß+F�:; , ��;; -
�*; ≡ �*;: + �*;;
Technology: �Z = ��*: , F*;:
Market: �Z ≡ _:: + �Z:; + #
�Z; = ß+F*:; , �*;; -
�Z; ≡ �Z;: + _;: + #;
Consumption
Consumption
Fig. 1. Structure of production and trade
Upstream: Invest to
accumulate capital and
hire labor to produce
intermediates
Midstream: Combine
domestic and imported
foreign intermediates to
produce intermediates
Downstream: Combine
domestic and imported
foreign intermediates to
produce final goods for
consumption and
investment
34
Fig. 2. Cross correlations
-20 0 20-0.4
-0.2
0
0.2
0.4
0.6GDP
-20 0 20-0.4
-0.2
0
0.2
0.4
0.6Consumption
-20 0 20-0.5
0
0.5
1Export
-20 0 20-0.5
0
0.5
1Import
-20 0 20-0.4
-0.2
0
0.2
0.4
0.6GDP
-20 0 20-0.4
-0.2
0
0.2
0.4Consumption
-20 0 20-0.5
0
0.5
1Export
-20 0 20-0.5
0
0.5Import
-20 0 20-0.4
-0.2
0
0.2
0.4
0.6GDP
-20 0 20-0.4
-0.2
0
0.2
0.4Consumption
-20 0 20-0.5
0
0.5
1Export
-20 0 20-0.5
0
0.5
1Import
Data Model
SEA4-EA5
CN-EA5
CN-SEA4
35
x-axis denotes period; y-axis indicates responses in percentage
Fig. 3. Dynamic response of GDP towards common and idiosyncratic positive TFP shock.
0 5 10 15 200.1
0.15
0.2
0.25
0.3
0.35
CN
0 5 10 15 20-0.1
0
0.1
0.2
0.3
0.4
EA
0 5 10 15 20-0.1
0
0.1
0.2
0.3
0.4
0 5 10 15 20-0.2
-0.1
0
0.1
0.2
0 5 10 15 20-0.1
0
0.1
0.2
0.3
SE
A
CN
0 5 10 15 20-0.15
-0.1
-0.05
0
0.05
0 5 10 15 200
0.05
0.1
0.15
0.2
0.25
SEA
0 5 10 15 20-0.2
-0.1
0
0.1
0.2
0 5 10 15 20-0.1
-0.05
0
0.05
0.1
EA
Common shock Idiosyncratic shock
Response
Shock
36
x-axis denotes period; y-axis indicates responses in percentage
Fig. 4. Dynamic response of GDP towards common and idiosyncratic negative preference
shock.
0 5 10 15 20-0.1
-0.05
0
0.05
0.1
CN
0 5 10 15 20-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
EA
0 5 10 15 20-0.1
-0.08
-0.06
-0.04
-0.02
0
0 5 10 15 20-0.1
-0.05
0
0.05
0.1
0 5 10 15 20-0.3
-0.2
-0.1
0
0.1
SE
A
CN
0 5 10 15 20-0.15
-0.1
-0.05
0
0.05
0.1
0 5 10 15 20-0.3
-0.2
-0.1
0
0.1
SEA
0 5 10 15 20-1.5
-1
-0.5
0
0.5
0 5 10 15 20-0.4
-0.2
0
0.2
0.4
EA
Common shock Idiosyncratic shock
Shock
Response